Closed-loop Artificial Pancreas: Algorithm Engineering and Clinical Evaluation

描述（由申请人提供）：该提案旨在建立在一个有前途的研究基础上，该研究已经产生了一个闭环人工胰腺（AP）的工作原型，该原型将连续葡萄糖传感与通过控制算法自动胰岛素输送相结合。AP由三个组件组成：持续葡萄糖传感器、持续皮下胰岛素输注泵和人工胰腺软件（APS）。美国食品药品监督管理局正在对APS进行最后阶段的审查，以用于全自动闭环控制临床试验（主文件MAF-1625）。具体目标是：1）开发人工胰腺以优化胰岛素输送并最小化与膳食相关的葡萄糖波动，2）扩展人工胰腺以包括胰岛素敏感性的改变和个体差异，以及3）进一步开发使用多参数模型预测控制（mpMPC）的算法以减少在线计算，制定有效的在线监控策略，以确保AP正常运行，并适应多种输入系统。为了实现我们的胰岛素输送全自动闭环装置的总体目标，我们将采用分阶段方法，其中临床研究也采用并行工程设计方法，以完善我们的AP。计算机模拟测试和人体临床测试将用于验证模型开发的每个步骤。这一系列研究的最终目标是功能性AP，其将通过响应于检测到的葡萄糖水平变化模式的受控胰岛素递送来提供全天候的葡萄糖调节，以便在1型糖尿病受试者中实现最佳的葡萄糖调节。在设置后续胰岛素输注的限制条件时，将特别考虑既往输注（体内胰岛素）中潜在可用的胰岛素量。该提案描述了血糖调节的多参数模型预测控制方法，以及AP在过夜、用餐和运动条件下的广泛计算机模拟和临床验证。该应用程序源于加州大学圣巴巴拉分校（UCSB）的系统工程师与著名的Sansum糖尿病研究所（Sansum）专门从事糖尿病研究的研究医生之间的长期合作，该研究所位于不到10英里的地方。该团队是人工胰腺文献的主要贡献者，也是糖尿病技术的国际领导者。 公共卫生相关性：该项目旨在开发一种功能性人工胰腺，该胰腺将通过控制胰岛素输送来响应检测到的葡萄糖水平变化模式，提供全天候的葡萄糖调节，以实现1型糖尿病受试者的最佳葡萄糖调节。在设置后续胰岛素输注的限制条件时，将特别考虑既往输注（体内胰岛素）中潜在可用的胰岛素量。该提案描述了血糖调节的多参数模型预测控制方法，以及AP在过夜、用餐和运动条件下的广泛计算机模拟和临床验证。

DESCRIPTION (provided by applicant): This proposal seeks to build on a promising line of research that has already resulted in a working prototype of a closed-loop artificial pancreas (AP) that combines continuous glucose sensing with automated insulin delivery via a control algorithm. The AP consists of three components: a continuous glucose sensor, a continuous subcutaneous insulin infusion pump, and artificial pancreas software (APS). The APS is in the final stages of review by the Food and Drug Administration for use in fully automated closed-loop control clinical trials (Master File MAF-1625 ). The specific aims are: 1) Development of the Artificial Pancreas to optimize insulin delivery and minimize meal-related excursions in glucose, 2) Expansion of the Artificial Pancreas to include alterations and individual differences in insulin sensitivity, and 3) Further development of the algorithms using multi-parametric model predictive control (mpMPC) in order to reduce online computation, develop effective online monitoring strategies to ensure that the AP is operating properly, and accommodate multiple input systems. In the quest to achieve our overall goal of a completely automated closed-loop device for insulin delivery, we will utilize a staged approach in which the clinical studies also have a concurrent engineering design approach in order to refine our AP. In silico testing and human in-clinic testing will be used to validate each step of model development. The ultimate goal of this line of research is a functional AP that will provide around-the-clock glucose regulation through controlled insulin delivery in response to detected patterns of changes in glucose levels in order to achieve optimal glucose regulation in subjects with type 1 diabetes. Special consideration will be given to the amount of potentially available insulin from prior infusions (insulin-on-board) in setting constraints for subsequent insulin delivery. The proposal describes multi-parametric model predictive control approaches to glycemic regulation and extensive in silico and clinical validation of the AP under overnight, meal, and exercise conditions. This application grows out of a long-standing collaboration between systems engineers at the University of California, Santa Barbara (UCSB) and research physicians specializing in diabetes research at the prestigious Sansum Diabetes Research Institute (Sansum) located less than ten miles away. This team has distinguished itself as a major contributor to the artificial pancreas literature and as an international leader in diabetes technology. PUBLIC HEALTH RELEVANCE: This project aims to develop a functional artificial pancreas that will provide around-the-clock glucose regulation through controlled insulin delivery in response to detected patterns of changes in glucose levels in order to achieve optimal glucose regulation in subjects with type 1 diabetes. Special consideration will be given to the amount of potentially available insulin from prior infusions (insulin-on-board) in setting constraints for subsequent insulin delivery. The proposal describes multi-parametric model predictive control approaches to glycemic regulation and extensive in silico and clinical validation of the AP under overnight, meal, and exercise conditions.